SkDebugCanvas incremental draw optimization

Optimization: SkDebugCanvas is capable of pausing and drawing incrementally instead of redrawing all commands.

Committed: https://code.google.com/p/skia/source/detail?r=4891

[DerekS, 2012-08-01 12:41:39]

http://codereview.appspot.com/6458056/diff/1008/debugger/QT/SkCanvasWidget.h
File debugger/QT/SkCanvasWidget.h (right):

http://codereview.appspot.com/6458056/diff/1008/debugger/QT/SkCanvasWidget.h#...
debugger/QT/SkCanvasWidget.h:121: SkIPoint fTranslate;
fOffset?

http://codereview.appspot.com/6458056/diff/1008/debugger/QT/SkGLWidget.cpp
File debugger/QT/SkGLWidget.cpp (right):

http://codereview.appspot.com/6458056/diff/1008/debugger/QT/SkGLWidget.cpp#ne...
debugger/QT/SkGLWidget.cpp:56: // TODO(chudy): Implement an optional flush
button in Gui.
I don't think that is necessary as this should be the correct behavior

http://codereview.appspot.com/6458056/diff/1008/debugger/QT/SkRasterWidget.cpp
File debugger/QT/SkRasterWidget.cpp (right):

http://codereview.appspot.com/6458056/diff/1008/debugger/QT/SkRasterWidget.cp...
debugger/QT/SkRasterWidget.cpp:44: fClip = fCanvas->getTotalClip().getBounds();
do you still need to store fMatrix and fClip in this widget?

http://codereview.appspot.com/6458056/diff/1008/debugger/SkDebugCanvas.cpp
File debugger/SkDebugCanvas.cpp (right):

http://codereview.appspot.com/6458056/diff/1008/debugger/SkDebugCanvas.cpp#ne...
debugger/SkDebugCanvas.cpp:91: }
move this logic to it's own method that you can share with getCommandPointAt

http://codereview.appspot.com/6458056/diff/1008/debugger/SkDebugCanvas.h
File debugger/SkDebugCanvas.h (right):

http://codereview.appspot.com/6458056/diff/1008/debugger/SkDebugCanvas.h#newc...
debugger/SkDebugCanvas.h:193: SkIPoint fTransform;
lets call it setOffset and fOffset as translate and transform have a slightly
different meaning.

[chudy, 2012-08-01 13:35:14]

http://codereview.appspot.com/6458056/diff/1008/debugger/QT/SkCanvasWidget.h
File debugger/QT/SkCanvasWidget.h (right):

http://codereview.appspot.com/6458056/diff/1008/debugger/QT/SkCanvasWidget.h#...
debugger/QT/SkCanvasWidget.h:121: SkIPoint fTranslate;
On 2012/08/01 12:41:39, DerekS wrote:
> fOffset?

Done.

http://codereview.appspot.com/6458056/diff/1008/debugger/QT/SkGLWidget.cpp
File debugger/QT/SkGLWidget.cpp (right):

http://codereview.appspot.com/6458056/diff/1008/debugger/QT/SkGLWidget.cpp#ne...
debugger/QT/SkGLWidget.cpp:56: // TODO(chudy): Implement an optional flush
button in Gui.
On 2012/08/01 12:41:39, DerekS wrote:
> I don't think that is necessary as this should be the correct behavior

We can talk about this, because of this flush call the GL widget looks different
at different commands than the raster. For example it pushes everything out the
save / restore layer instead of waiting till restore is called

http://codereview.appspot.com/6458056/diff/1008/debugger/QT/SkRasterWidget.cpp
File debugger/QT/SkRasterWidget.cpp (right):

http://codereview.appspot.com/6458056/diff/1008/debugger/QT/SkRasterWidget.cp...
debugger/QT/SkRasterWidget.cpp:44: fClip = fCanvas->getTotalClip().getBounds();
On 2012/08/01 12:41:39, DerekS wrote:
> do you still need to store fMatrix and fClip in this widget?

This will be part of the refactoring process. We have to figure out someway to
calculate it in SkDebugCanvas. Added TODO

http://codereview.appspot.com/6458056/diff/1008/debugger/SkDebugCanvas.cpp
File debugger/SkDebugCanvas.cpp (right):

http://codereview.appspot.com/6458056/diff/1008/debugger/SkDebugCanvas.cpp#ne...
debugger/SkDebugCanvas.cpp:91: }
On 2012/08/01 12:41:39, DerekS wrote:
> move this logic to it's own method that you can share with getCommandPointAt

Done.

http://codereview.appspot.com/6458056/diff/1008/debugger/SkDebugCanvas.h
File debugger/SkDebugCanvas.h (right):

http://codereview.appspot.com/6458056/diff/1008/debugger/SkDebugCanvas.h#newc...
debugger/SkDebugCanvas.h:193: SkIPoint fTransform;
On 2012/08/01 12:41:39, DerekS wrote:
> lets call it setOffset and fOffset as translate and transform have a slightly
> different meaning.

Done.

[DerekS, 2012-08-01 15:54:46]

lgtm with one nit

http://codereview.appspot.com/6458056/diff/8001/debugger/SkDebugCanvas.h
File debugger/SkDebugCanvas.h (right):

http://codereview.appspot.com/6458056/diff/8001/debugger/SkDebugCanvas.h#newc...
debugger/SkDebugCanvas.h:47: void applyUserTransform(SkCanvas* canvas);
should this be private?

[chudy, 2012-08-01 15:58:18]

Committed in revision 4891. Thanks for the review!

http://codereview.appspot.com/6458056/diff/8001/debugger/SkDebugCanvas.h
File debugger/SkDebugCanvas.h (right):

http://codereview.appspot.com/6458056/diff/8001/debugger/SkDebugCanvas.h#newc...
debugger/SkDebugCanvas.h:47: void applyUserTransform(SkCanvas* canvas);
On 2012/08/01 15:54:46, DerekS wrote:
> should this be private?

Done. Moved.